SOFTWARE-RELIABILITY MODEL WITH OPTIMAL SELECTION OF FAILURE DATA

In the use of software reliability models it is not necessarily the ca se that all the failure data should be used to estimate model paramete rs and to predict failures. The reason for this is that old data may n ot be as representative of the current and future failure process as r ecent data. Therefore, it may be possible to obtain more accurate pred ictions of future failures by excluding or giving lower weight to the earlier failure counts. Although ''data aging'' techniques such as mov ing average and exponential smoothing are frequently used in other fie lds, such as inventory control, we did not find use of data aging in t he various models we surveyed. One model that includes the concept of selecting a subset of the failure data is the Schneidewind Non-Homogen eous Poisson Process (NHPP) software reliability model. In order to us e the concept of data aging, there must be a criterion for determining the optimal value of the starting failure count interval. We evaluate d four criteria for identifying the optimal starting interval for esti mating model parameters. Three of the criteria are novel. Two of these treat the failure count interval index as a parameter by substituting model functions for data vectors and optimizing on functions obtained from maximum likelihood estimation techniques. The third one uses wei ghted least squares to maintain constant variance in the presence of t he decreasing failure rate assumed by the model. The fourth criterion is the familiar mean square error. Our research showed that significan tly improved reliability predictions can be obtained by using a subset of the failure data, based on applying the appropriate criteria, and using the Space Shuttle On-Board software as an example.